AAV6 vectors provide superior gene transfer compared to AAV9 vectors following intramyocardial administration

Abstract

Cardiac gene therapy using adeno-associated viral (AAV) vectors holds great promise for treating heart diseases but would benefit from more potent AAV vectors. Vectors based on the AAV serotypes 6 and 9 have been used in pre-clinical gene therapy studies, yet the therapeutic outcomes varied depending on the experimental model and delivery route used. Here, we evaluated the transduction efficiency of AAV6, AAV9, and AAV9-derived MyoAAVs for local cardiac delivery. Vectors were tested in neonatal rat ventricular myocytes, and subsequently in mouse hearts by direct intramyocardial injection. Vector genome levels, mRNA expression levels, and fluorescence were measured. The AAV6 and AAV9 vectors were further validated in porcine hearts, human-induced pluripotent stem-cell-derived cardiomyocytes, and human atrial myocardial slices. In both rat cardiomyocytes and mouse hearts, AAV6 exhibited the highest transduction efficiency. Direct comparison of the AAV6 and AAV9 vectors in porcine and human models confirmed that AAV6 is more potent. In conclusion, AAV6 vectors are superior to AAV9 and its derivative vectors for cardiac transduction by direct intramyocardial injection. In addition, the in vivo transduction efficiency correlates with in vitro and ex vivo assays, thereby facilitating the development of more potent AAV variants for cardio-selective delivery methods.

Graphical abstract

Boink and colleagues demonstrate that AAV6 provides superior transduction efficiency over AAV9 and MyoAAV4A in locally applied gene therapy in vitro, ex vivo, and in vivo in various rodent, pig, and human models. These findings underscore the robust transduction efficiency with AAV6 in the setting of direct intramyocardial injection.

Introduction

Cardiovascular disease remains the top health threat in terms of mortality and morbidity, affecting 30% of the global population.¹ Gene therapy is a promising method to treat diseases by delivering genetic information into target cells, ultimately altering cellular function to improve disease status. Adeno-associated viral (AAV) vectors have become an attractive delivery vehicle for in vivo gene delivery purposes due to their ability to achieve long-term transgene expression and their low immunogenicity.² To date, more than 350 clinical trials have been carried out employing AAV vectors,³ and a total of six AAV-based gene therapy products (Glybera, Luxturna, Zolgensma, Roctavian, Hemgenix, and Kebilidi/Upstaza) have received regulatory approval from either the Food and Drug Administration (FDA) or the European Medicines Agency (EMA).^4,5 However, the development of gene therapies for cardiovascular diseases lags behind that of other organ systems. No AAV-based cardiac gene therapy has been approved yet, partially due to the lack of effective AAV gene delivery methods in large mammalian (including human) hearts.^6,7 Failure of the largest cardiac clinical gene therapy trials performed to date, which utilized AAV1 vectors, has largely been attributed to inefficient gene transfer.^8,9 As a result, recent and ongoing clinical trials have shifted toward alternative AAV serotypes in an effort to enhance gene delivery efficacy.^{10,11,12,13,14,15}

Hundreds of AAV serotypes have been discovered in humans, non-human primates, and various other species, displaying different tissue and cell type tropisms.¹⁶ Among these serotypes, AAV1, AAV6, AAV8, AAV9, AAVrh10, and AAVrh74 have been demonstrated to be cardiotropic in different settings.^{17,18,19,20,21,22,23} Among these cardiotropic serotypes, AAV6 and AAV9 are the most widely used and arguably the most potent for cardiac gene transfer.^{24,25,26,27,28} In various in vitro and ex vivo models, including neonatal rat ventricular cardiomyocytes (NRVMs), rat and pig cardiac slices, human embryonic stem-cell-derived cardiomyocytes, human-induced pluripotent stem-cell-derived cardiomyocytes (hiPSC-CMs), and human myocardial slices, AAV6 outperformed AAV9.^29,30 On the other hand, in several other comparative cardiac gene therapy studies, AAV9 performed equally well as, or even outperformed, AAV6.^{3,24,30,31,32,33} This led to the notion that AAV transduction properties in vitro/ex vivo may differ from those in vivo. Importantly, most in vivo studies were done after intravenous vector delivery, which may not translate well to other routes of vector administration.

Intramyocardial injection represents a highly attractive method for attaining high transduction efficiency within a defined target region. It is particularly useful in gene-therapy-mediated biological pacemakers, where focal delivery of the vectors is critical to create a local pacemaker substrate while leaving the rest of the myocardium unaffected.³⁴ Moreover, this mode of gene delivery is also considered for the treatment of myocardial infarction and scarring, aiming to promote local tissue regeneration^35,36,37,38 or prevent ventricular arrhythmias,^39,40 which in case of the former is now progressing toward the first-in-human clinical trial testing.¹⁰ Another advantage of intramyocardial injection is the reduced exposure in the bloodstream, which significantly minimizes the risk of neutralization by circulating antibodies.⁴¹ Intramyocardial injection can be performed either surgically or percutaneously.^42,43 The surgical approach, involving open-heart surgery, offers high precision and flexibility in vector delivery but is invasive. In contrast, the percutaneous approach uses catheter-based transendocardial injection, making it minimally invasive and thus more favorable in clinical settings.

Unlike studies employing intravenous injection, comparative studies using intramyocardial injection are limited and inconsistent. In these studies employing intramyocardial injection, AAV6 was reported the most potent in pigs, dogs, and non-human primates, suggesting it is superior to AAV9.^35,44,45 In contrast, a mouse study showed AAV6 to be inferior to AAV9 in cardiac gene transfer upon intramyocardial injection.⁴⁶ Moreover, intramyocardial injection of AAV1-8 vectors in rats showed AAV8 to be better than AAV6,⁴⁷ suggesting a suboptimal performance of AAV6 in cardiac gene transfer. Such profound difference in AAV serotype performance between different models complicates the development of more potent AAV variants. It potentially also contributed to the selection of suboptimal serotypes and the application of unnecessarily high vector dosages for pre-clinical and clinical studies.^{20,38,48,49,50,51,52,53} Ultimately, such suboptimal serotypes require (much) higher dosages, leading to increased risk of side effects and immune activation as well as higher production costs.

Our present work systematically evaluates the performance of different AAV serotypes in terms of cardiac transgene expression and transduction efficiency utilizing various in vitro, in vivo, and ex vivo models. Our results indicate that, for AAV vector-mediated gene transfer to cardiomyocytes, in vitro and ex vivo results obtained with a particular AAV vector are highly predictive for the in vivo transduction efficiency via direct injection and that the performance of AAV vectors in different species is consistent. Moreover, these experiments confirm the potency of AAV6 in cardiac gene transfer across different species, including mouse, pig, and human, particularly in the setting of direct intramyocardial injection.

Results

AAV6 and MyoAAV4A achieve high transgene expression in NRVMs

We started by producing six different AAV vectors carrying a chicken Tnnt2 (cTnT) promoter-driven enhanced green fluorescent protein (GFP) transgene (Figure 1A) and pseudotyped with capsids of two natural cardiotropic AAV serotypes (i.e., AAV6 and AAV9) and four engineered AAV9 capsids selected for high muscular transduction (MyoAAV2A, -3F, -4A, and -4E).²⁶ The production yields of these vectors were similar, ranging from 3.3 to 5.3 × 10¹³ vector genomes (vg) (Figure S1A). To validate the performance of the engineered AAV capsids, we injected mice intravenously with MyoAAV4A and AAV9 and found that, similar to previous reports,^26,54 MyoAAV4A led to significantly higher transgene expression compared to AAV9 4 weeks post-injection (Figures S1B and S1C).

Figure 1.

AAV6 and MyoAAV4A efficiently transduce NRVMs

(A) Experimental design. (B) Direct fluorescent images of NRVMs 3 days post-transduction with various AAV vector pseudotypes. Scale bars, 100 μm. (C) Percentage of GFP-expressing cells determined by flow cytometry. (D) Relative expression levels of GFP determined by RT-qPCR. Data are presented as mean ± SEM. Data were compared using one-way ANOVA with post-hoc Fisher’s LSD test. ∗∗p < 0.01; ∗∗∗, ###p < 0.001. ∗ denotes comparison between groups and UT. # denotes comparison between groups and AAV6. UT, untransduced; AAV9_10×, AAV9 at a 10 times higher dose.

We subsequently compared the transgene expression in vitro in NRVMs transduced with all of the above-mentioned AAV vectors at an MOI of 10⁴ vg/cell and incubated for 3 days until flow cytometry and reverse-transcription quantitative PCR (RT-qPCR) analyses (Figures 1A and 1B). Additionally, a 10-fold higher dose for AAV9 (AAV9_10×) was included to explore the effect of differential vector dosing. Flow cytometry revealed that AAV6 leads to the highest transduction efficiency in NRVMs, followed by MyoAAV4A and MyoAAV2A (Figure 1C). AAV9 and AAV9_10× gave the lowest transduction efficiency of all AAV vector pseudotypes, significantly lower than AAV6 (Figure 1C). RT-qPCR analyses revealed a similar pattern as flow cytometry, with AAV6 giving the highest transgene expression, followed by MyoAAV4A and MyoAAV2A (Figure 1D). From these results, we selected the most potent MyoAAV variant MyoAAV4A, as well as AAV6 and AAV9 for further in vivo testing.

AAV6 outperforms AAV9 and MyoAAV4A in direct intramyocardial-injection-mediated gene transfer in mouse myocardium

Subsequently, we compared the ability of the cTnT-driven AAV6, AAV9, and MyoAAV4A pseudotype vectors to transduce healthy mouse myocardium after direct intramyocardial injection. We injected 10¹¹ vg of these vectors into the left ventricular wall of female FVB mice (Figure 2A). An additional high-dose (10¹² vg) group (AAV9_10×) for AAV9 was included. Four weeks later, GFP fluorescence was observed in all AAV-vector-injected hearts using direct fluorescence imaging (Figure 2B), further microscopically confirmed by immunofluorescence staining of GFP at the injection site (Figure 2C). To evaluate transgene expression levels, we utilized macroscopy for fluorescence quantification and RT-qPCR for mRNA quantification. Both methods revealed AAV6 as the most efficient variant for transgene expression, followed by MyoAAV4A (Figures 2D and 2E). Similar to our in vitro results (Figures 1C and 1D), AAV9_10× led to significantly higher transgene expression compared to AAV9, but the expression levels still remained significantly lower than those of AAV6 or MyoAAV4A (Figures 2D and 2E). To estimate transduction efficiency, the vector copy number was quantified, showing AAV6 to be the most efficient of the tested AAV vectors (Figure 2F). Both fluorescence and mRNA expression showed significant correlation to vector genome copy number (Figure S2). None of the tested vectors induced hepatic transgene expression, and vector genomes were detected only in the liver of animals receiving the high dose of AAV9 and one AAV6 animal (Figure S3). Our data indicate that AAV6 is more efficient in transducing mouse cardiomyocytes than AAV9 or MyoAAV4A, which is in line with the results in NRVMs (Figure 1).

Figure 2.

AAV6 and MyoAAV4A efficiently transduce mouse myocardium

(A) Experimental design. (B) Direct fluorescent images of mouse hearts. Scale bars, 2 mm. (C) Immunostaining images of hearts and zoom-in images from the injection sites. Scale bars, 1 mm (upper) or 200 μm (lower). (D) Quantification of the integrated density of direct GFP fluorescence, n = 6. (E) mRNA expression level of GFP in left ventricles, n = 3. (F) AAV vector genome copies in ventricles, n = 3. Data are presented as mean ± SEM. Data were compared using one-way ANOVA with post-hoc Fisher’s LSD test. ∗, #p < 0.05; ∗∗, ##p < 0.01; ∗∗∗, ###p < 0.001. ∗ denotes comparison between groups and PBS. # denotes comparison between groups and AAV6. IM, intramyocardial; AAV9_10×, AAV9 at a 10 times higher dose; vg/dg, vector genomes per diploid genome.

Intramyocardial injection of AAV vectors in mice does not induce vector-related cardiac fibrosis at 4 weeks post-injection

To assess the potential toxicity of intramyocardial AAV injections, we quantified cardiac fibrosis using histological staining on heart sections from Dulbecco’s PBS- and AAV vector-injected mouse hearts (Figure 3). Picrosirius red staining revealed moderately increased collagen levels, indicating mild cardiac fibrosis at the injection site in both PBS- and AAV-treated animals (Figure 3A). No significant differences were observed among the groups including the control PBS group (Figure 3B), suggesting that the fibrosis might stem from the injection procedure itself rather than from the AAV transduction, for example due to needle track trauma or puncture of the epicardium. Additionally, consecutive sections stained for collagen and GFP showed no correlation between the extent of fibrosis and GFP expression (Figure 3C). These results indicate that AAV vector transduction and GFP expression do not increase cardiac fibrosis in hearts following intramyocardial injection at least within the first 4 weeks post-injection.

Figure 3.

Intramyocardial injection of AAV vector does not induce vector-related cardiac fibrosis

(A) Example of picrosirius red staining images of hearts injected with PBS or AAV6-cTnT-GFP vector. Black arrows indicate regions of cardiac fibrosis. Scale bars, 1 mm. (B) Quantification of the fibrosis area in hearts injected with PBS or AAV vectors. No significant differences were observed among the groups, n = 3. (C) Picrosirius red (left) and immunofluorescence (right) staining images of a heart injected with AAV6-cTnT-GFP. Stainings were performed using consecutive sections. Scale bars, 500 μm. Data are presented as mean ± SEM. Data were compared using one-way ANOVA with post-hoc Fisher’s LSD test. ns, not significant.

AAV6 more efficiently transduces mouse myocardium after intramyocardial injection than AAV9 independently of transgene promoter and mouse sex or strain

Our results have shown that AAV6 is more efficient than AAV9 in cardiac gene transfer in vitro and in vivo upon direct intramyocardial injection (Figures 1 and 2). However, these results were performed exclusively with vectors produced in Amsterdam UMC using the cardiomyocyte-specific cTnT promoter. To test the robustness of our findings, we repeated part of the experiments using AAV vectors carrying a ubiquitous human cytomegalovirus (CMV) immediate-early gene promoter, produced and titrated by Revvity Gene Delivery, and we included mice from different strains and sex (FVB female and C57BL/6J male) (Figure 4), as these variables are all known to potentially impact AAV performance.^{55,56,57,58,59,60}

Figure 4.

AAV6 outperforms AAV9 independently of the expression cassette and mouse background

(A) Scheme of the in vitro experiment. (B) Direct fluorescent images of NRVMs 3 days post-transduction with AAV6- and AAV9-pseudotyped vectors. Yellow arrows indicate weakly GFP-expressing cells. Scale bars, 100 μm. (C) Expression level of GFP mRNA determined by RT-qPCR, n = 4. (D) Scheme of the in vivo experiment and representative heart images. Scale bars, 2 mm. (E) Quantification of direct GFP fluorescence, (F) GFP mRNA expression, and (G) AAV vector genome copies in FVB female mice injected with AAV6- or AAV9-CMV-GFP, n = 3. (H) Scheme of the in vivo experiment and representative heart images. Scale bars, 2 mm. (I) Quantification of direct GFP fluorescence, (J) GFP mRNA expression, and (K) AAV vector genome copies in C57BL/6J male mice injected with AAV6- or AAV9-CMV-GFP, n = 4. Data are presented as mean ± SEM. Data were compared using Student’s t test. ∗p < 0.05; ∗∗p < 0.01. ∗ denotes comparison between AAV6 and AAV9. vg/dg, vector genomes per diploid genome.

First, we transduced NRVMs with AAV6 and AAV9 pseudotype vectors in which GFP expression was driven by the CMV promoter and quantified GFP-specific mRNA levels three days post-transduction (Figure 4A). GFP expression was detected by direct fluorescence microscope in both groups (Figure 4B), although AAV6 led to significantly higher transgene expression in NRVMs compared to AAV9 (Figure 4C). These findings are in line with our results obtained with the cTnT-promoter-containing AAV vectors produced in Amsterdam UMC (Figure 2).

We then injected these vectors into the left ventricular free wall of FVB female (as used in Figure 2) and C57BL/6J male mice, which were previously used by Prasad et al.⁴⁶ to compare the gene transfer efficiency of different AAV vector pseudotypes following intramyocardial injection. In FVB female mice, AAV6 resulted in a 9.4-fold higher GFP mRNA level than AAV9 while the GFP fluorescence (p = 0.11) and vector copy number (p = 0.10) did not significantly differ between the two vectors (Figures 4D–4G). Also in C57BL/6J male mice, AAV6 exhibited a higher transduction efficiency than AAV9 (Figures 4H–4K). We next examined transgene expression in the liver of the animals (Figure S4). Hepatic transgene expression visualized by direct GFP fluorescence was only detected in C57BL/6J males (Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Figure 6S4A–S4C), in accordance with a previous report indicating sex differences in liver transduction.⁶¹ Additionally, AAV9 led to significantly more gene expression and fluorescence than AAV6 (Figures S4C and S4D). Collectively, our data suggest that AAV6 is more efficient than AAV9 in cardiac gene transfer independently of the expression cassette and mouse sex or strain, while also leading to lower hepatic transduction.

Figure 5.

AAV6 outperforms AAV9 in transducing pig myocardium

(A) Experimental design. (B) Direct fluorescent macroscopy of pig heart tissues injected with AAV6- or AAV9-CMV-GFP. Sampled tissues are encircled using a white dashed line. Scale bars, 1 cm. (C) Immunostaining images from pig heart tissues injected with 200 μL (high dose) AAV6- or AAV9-CMV-GFP. Scale bars, 200 μm. (D) Quantification of the GFP mRNA expression and (E) AAV vector genome copies in pig heart tissues injected with AAV6- or AAV9-CMV-GFP, n = 3. Data are presented as mean ± SEM. Data were compared using two-way ANOVA with post-hoc Fisher’s LSD test. ∗,#p < 0.05; ##p < 0.01; ∗∗∗p < 0.001. ∗ denotes comparison between AAV serotypes. # denotes comparison between injection volumes. IM, intramyocardial. vg/dg, vector genomes per diploid genome.

Figure 6.

AAV6 outperforms AAV9 in transducing hiPSC-CMs and human heart slices

(A) Scheme of the hiPSC-CM experiment. (B) Quantification of the AAV vector genomes in hiPSC-CMs transduced with AAV6- or AAV9-CMV-GFP, n = 4. (C) Scheme of the human LAAS experiment. (D) Quantification of the AAV vector genomes in human LAASs transduced with AAV6- or AAV9-CMV-GFP, n = 4 donors. Data are presented as mean ± SEM. Data were compared using two-way ANOVA with post-hoc Fisher’s LSD test (B) or Student’s t test (D). ∗p < 0.05; ∗∗∗, ###p < 0.001. ∗ denotes comparison between AAV serotypes. # denotes comparison between multiplicities of infection. MOI, multiplicity of infection. vg/dg, vector genomes per diploid genome.

AAV6 is more efficient in direct cardiac gene transfer than AAV9 in pig myocardium

To test if there is a species-dependent transduction efficiency, we compared the transduction efficiency of AAV6 and AAV9 in pigs, a relevant translational large animal model for cardiac diseases. Male pigs were injected at low (10¹¹ vg in 20 μL) or high dose (10¹² vg in 200 μL) with CMV-GFP AAV6 and AAV9 vectors that were applied via sub-epicardial injection at two separate sites in the left ventricle (Figure 5A). Two weeks post-injection, hearts were exposed, and injection sites were assessed by macroscopic GFP fluorescence and suture marks (Figure S5). At each injection site, direct GFP fluorescence was detected in the surrounding myocardium (Figure 5B). GFP expression was further confirmed via immunofluorescence microscopy (Figure 5C). RT-qPCR again showed that AAV6 led to significantly higher transgene expression and more vector genomes at the high-dose injection site, when compared to AAV9 (Figures 5D and 5E). Similar trends were observed at the low-dose injection sites (Figures 5D and 5E), although they did not reach statistical significance. These samples exhibited greater variability, likely due to imprecise sampling of the small, transduced area, which resulted in more tissue outside the injection site being included in the analyzed tissue samples. Altogether, these data suggest that AAV6 is more efficient in transducing pig myocardium than AAV9 upon direct intramyocardial injection.

AAV6 outperforms AAV9 in transducing hiPSC-CMs and human myocardial slices

Next, we investigated whether we could recapitulate the findings from the rodent and pig experiments in human translational models. We transduced hiPSC-CMs at MOIs of 2 × 10³ and 2 × 10⁴ vg/cell with AAV6 and AAV9 pseudotype vectors carrying the CMV promoter-driven GFP transgene (Figure 6A). Five days later, GFP fluorescence was observed in all conditions, and cells were collected for further molecular analyses. RT-qPCR on AAV vector genomes showed a significantly higher transduction efficiency of AAV6 compared to AAV9 at both MOIs (Figure 6B). These results are in accordance with previous reports.^29,62 We continued our study with an established ex vivo human left atrial appendage slices (LAAS) model. Human LAASs were transduced with 10¹¹ vg vectors and incubated for 3 days before molecular analyses (Figure 6C). RT-qPCR on AAV vector genomes revealed that AAV6 more efficiently transduced the LAASs as compared to AAV9 (Figure 6D). Collectively, these data suggest that AAV6 outperforms AAV9 in transducing human myocardium.

Discussion

Our data indicate that AAV6 pseudotyped vectors outperform AAV9 pseudotyped vectors, a benchmark in the cardiac gene therapy field, and AAV9-derived engineered variants, regarding cardiac transduction efficiency following intramyocardial injection.

Both AAV6 and AAV9 have been successfully utilized in cardiac gene therapy studies via intramyocardial injection, showing reasonable transduction efficiencies.^35,38 However, identifying more potent AAV vectors can further aid in achieving clinically relevant transduction levels with acceptable vector doses and help reduce inter-individual variability in cardiac transduction efficiencies.^6,7 In studies comparing AAV6 and AAV9 through intramyocardial injections, results have varied between small (mouse) and large animals (pig, dog, and non-human primates), sometimes showing conflicting outcomes.^44,45,46 In our study, we compared both vectors in cultures of rat and human cardiomyocytes, human atrial tissue slices, and mouse and pig hearts. We did not observe such discrepancies in our experiments, as AAV6 consistently outperformed AAV9 in terms of transgene expression and transduction efficiency. We also assessed the potency of AAV vectors in mice using two independent vector sources and two mouse strains and used different durations of transduction for in vitro, mouse and pig experiments to further validate the robustness of our findings. One explanation for the previously reported discrepancy between small and large animals is the use of a relatively large needle (29 G) in the mouse study.⁴⁶ This could generate more vector leakage into the circulation, for example by vessel puncture or back-flush, resulting in a different transduction environment where vectors that are more potent in the context of systemic administration, such as AAV9, may perform better.^24,31,32 In support of this hypothesis, Prasad et al. reported significantly higher transgene expression (up to 35-fold) in liver compared to controls, despite the use of the cardiomyocyte-specific cTnT promoter,⁴⁶ while we detected none when injecting vector with the same promoter using Hamilton syringes (31-G needles). As a result, our data highlight the robust consistency across different models and are very encouraging with regard to eventual application in human patients, proposing AAV6 as an efficient delivery vehicle for cardiac gene transfer in the context of cardio-selective vector delivery methods such as direct intramyocardial injection and likely also other local methods such as epicardial painting and intracoronary infusion.^63,64,65,66

Although AAV9 has a lower transduction efficiency than AAV6, it still performed reasonably well in our work, particularly when combined with the strong CMV promoter, which may explain the successful transduction observed in several studies using AAV9.^38,49,51,53 Additionally, AAV9 is currently being used in multiple clinical trials, utilizing intramyocardial, intracoronary, and intravenous injection methods^{10,11,12,13,14} and remains a promising tool for cardiac gene therapy. However, for applications requiring a high transduction efficiency, such as those based on dual AAV vectors or those requiring correction of a genetic defect in the majority of in vivo cardiomyocytes, AAV9 might not provide the required efficiency, potentially compromising functional outcomes. While administering higher doses can increase transduction efficiency, it also leads to more off-target transduction, as shown in our in vivo experiments. Additionally, excessively high doses may raise safety concerns, as vector doses correlate to vector immunogenicity in clinical trials,⁶⁷ and high doses reportedly influenced the contractility performance of cardiac slices.⁶⁸

In our NRVM and mouse experiments, we included AAV variants named MyoAAVs. These variants were generated by directed evolution of AAV9 toward muscular tropism via intravenous injection and exhibited stronger binding to cardiomyocytes through their RGD motifs.²⁶ To our knowledge, we are the first group to study the transduction efficiency of these variants in the context of intramyocardial injection. In mouse hearts, MyoAAV4A led to significantly higher transgene expression and transduction efficiency compared to its parental serotype AAV9 when intramyocardially injected. Despite not outperforming AAV6, it led to high transgene expression and could serve as a reasonable alternative to AAV6.

Dosing plays a critical role in gene therapy, as it influences both transduction efficiency and the immune response. The dose we used in mice was 10¹¹ vg/mouse (except for AAV9_10× group), which is around 2- to 10-fold higher than the previous mouse experiments.^46,48,49 This dose achieved high transgene expression in a large region of the heart in case of AAV6. However, direct comparisons of expression level and spread between this and previous studies are not possible due to the different sampling and detection methods. The dose we used in pigs was 10¹² vg per injection for the 200 μL injection site, which matches the dose reported by Gabisonia et al.³⁵ The transduction efficiency, as measured by viral genome copies, was also comparable to their findings.³⁵ At this dose, AAV6 administration resulted in 6.4 vg/dg at the injection site, indicating a high level of transduction. Multiple injections may be required to adequately cover the desired region but AAV6 at this dose appears to be sufficient to achieve effective transduction for each individual injection site. In this study, we used immunosuppressed pigs, and it is possible that in immunocompetent pigs the high local transduction evokes immune responses. It has been reported that intramyocardial injection of AAV2 expressing TNFRII-Fc at 5 × 10¹¹ vg per injection site (5 × 10¹² vg in total) triggered immune responses while the same dose of empty AAV2 did not.⁶⁹ This suggests that the immune response does not solely depend on the vector dose. Indeed, while the magnitude of the immune responses depends on the vector dose,⁶⁷ it is also affected by several other factors.⁷⁰ Therefore, immune responses must be carefully evaluated for each specific gene therapy, taking these variables into account.

To date, AAV engineering/screening efforts aiming at improving cardiac transduction have used intravenous injection.^26,71,72 However, these attempts always result in not only capsids with stronger binding properties to cardiomyocytes but also those with lower liver binding or better endothelial passage,^71,73,74 making it difficult to translate results to locally administered applications.⁷⁵ Local vector application, such as intramyocardial injection, intracoronary infusion, or epicardial painting, differs from intravenous injection in having minimal off-target vector sequestration and reduced necessity for endothelial passage. Moreover, different AAV serotypes utilize distinct primary receptors for cell attachment and entry. AAV6 primarily binds to sialic acid and heparan, while AAV9 primarily binds to terminal galactose and uses the laminin receptor.^29,76 These differences in vector sequestration, endothelial passage necessity, and receptor affinity likely together contribute to the different performance of AAVs in different delivery methods. We therefore expect more efficient AAV optimization using screening methods tailored to local vector application, and we expect that variant libraries based on AAV6 combined with these methods will yield further improved variants.

In the context of intramyocardial injections, in contrast to systemic applications, we propose in vitro and ex vivo assays we used may serve as predictors for vector performance, thereby providing useful platforms for optimization studies and validation in human tissue. The in vitro and ex vivo environments are lacking in complexity due to the absence of blood vessels and a competent immune system, allowing direct viral-to-cell contacts without competing vector absorption in off-target organs. These characteristics resemble the conditions of direct intramyocardial injection, where off-target transduction and contact with cellular and humoral immunity are minimized, and viral vectors do not need to pass the endothelial barrier.^41,77 Our results obtained with NRVMs closely predicted the outcomes of the mouse experiments. Additionally, comparison of natural AAV serotypes in various in vitro and ex vivo models has shown results largely consistent with those from in vivo experiments utilizing intramyocardial delivery,^29,62,68,78 further indicating high predictive power of these in vitro and ex vivo models.

Two limitations of this study should be acknowledged. First, our in vivo experiments were conducted up to 4 weeks post-injection, which may be insufficient to detect delayed changes in transgene expression or late-onset toxicity. However, AAV-mediated transgene expression is generally considered stable beyond 4 weeks post-injection,^24,25,79 making significant changes unlikely at later stages. Second, despite our best efforts, slight variability in the injection site location within the mouse hearts existed due to technical challenges. To mitigate this, we consistently sampled the lower half of the heart, ensuring inclusion of the injection site and maintaining comparable tissue composition across samples.

Our study showed that AAV6 is more efficient in transducing cardiac tissue than AAV9 in the context of direct intramyocardial injection. Moreover, our in vitro assays of cardiomyocyte transduction represent a sensitive and selective method to predict in vivo vector performance in this setting, thus representing useful tools for future AAV vector engineering efforts. These insights will facilitate the development of locally applied cardiac gene therapies and set the stage for subsequent optimization efforts and translational studies.

Materials and methods

Animal experiments

All animal experiments were approved and monitored by the Animal Ethics Committees of the UMC Utrecht and Amsterdam UMC and conducted in accordance with Directive 2010/63/EU of the European Parliament and with the “Guide for the Care and Use of Laboratory Animals.”

Mouse experiments

Eight-week-old FVB female mice and C57BL/6J male mice were acquired from Janvier (Le Genest-Saint-Isle, France) and acclimated for a week before the operation. Mice were injected subcutaneously with buprenorphine (0.075 mg/kg) and carprofen (0.05 mg/kg) for analgesia, at least 30 min prior to surgery. Anesthesia was induced with 4% isoflurane in 1 L/min O₂. Mice were shaved, intubated, and placed on a heating mat to maintain body temperature. Subsequently, an analgesic mixture consisting of lidocaine (2 mg/kg) and bupivacaine (3 mg/kg) was applied subcutaneously at the site of the incision. Anesthesia was maintained using ventilation with 1.5% isoflurane in 1 L/min O₂ mixed with 0.15 L/min air. Left thoracotomy was performed to expose the apex of the heart at the fourth intercostal space. To inject the AAV vectors into the apex, a 10 μL Hamilton syringe (RN1701) fitted with a 31-G needle (13 mm, point style 4) was inserted from the anterior left ventricle toward the apex in a shallow angle to prevent penetrating into the lumen. The viral vector solution was slowly administered in four injections of 5 μL at the same location to administer a total of 20 μL of viral vector (5×10¹² or 5×10¹³ vg/mL). The thoracotomy and skin were closed with a C-1 12-mm cutting needle with a 6-0 silicone-coated braided silk wire. Post-surgical analgesia consisted of four days of ad libitum carprofen (0.06 mg/mL) in drinking water and high caloric wet food.

Pig experiments

Four male Topigs Norsvin pigs (Van Beek SPF Varkens; weight 34.13 [6.67] kg) were conventionally housed in stables with hay bedding, day/night light cycles of 12 h, and fed with standard chow diet and water ad libitum. Three days before AAV injection, we initiated an oral immunosuppressive regimen consisting of cyclosporine A (100 mg/kg/day) and prednisolone (2 mg/kg/day) to avert an immune response against the viral vectors. One day before AAV injection, anticoagulation treatment was initiated (for consistency with other gene therapy studies in our laboratory) with carbasalate calcium (320 mg one day before surgery and 160 mg on the day of the surgery), and analgesia was achieved with a buprenorphine patch (20 μg/h). Both the immunosuppressive and the anticoagulation therapies were maintained until termination. On the day of the injection, animals were premedicated intramuscularly with ketamine (10 mg/kg), midazolam (0.4 mg/kg), and atropine (0.05 mg/kg). Anesthesia induction was achieved with thiopental (4 mg/kg), followed by endotracheal intubation and anesthesia maintenance with a constant-rate infusion of midazolam (0.4 mg/kg/h), sufentanil (2.5 μg/kg/h), and cis-atracurium (0.7 mg/kg/h). Pigs were mechanically ventilated with a positive pressure ventilator with FiO₂ 0.5, 1.5 L/min and an average frequency of 14 respirations/min under continuous capnography. Immunosuppressive treatment could not be assured orally during surgery; thus we performed a 2-h infusion of 200 mg in 100 mL of cyclosporine starting 1 hour before the injections. Pre-incisional analgesia was achieved with lidocaine (10 mg/mL), continued with a sternotomy and a full pericardiotomy to gain sub-epicardial access. Sternal homeostasis was assured using bone wax. Next, the heart was stabilized using a cardiac positioner (Starfish) ensuring hemodynamic stability, and three sutures (6-0 Prolene) were sewn to demarcate injection sites. We conducted four sub-epicardial injections using insulin syringes (BD Micro-Fine+), two of which contained 200 μL (high dose) of viral particle solution and two 20 μL (low dose) of viral particles solution (5 × 10¹² vg/mL), the volume of each being dispensed in 40 s. Later, two drains were inserted in the pleural and pericardial cavities, anchored with a purse-string suture (2-0 FS-1 Vicryl) followed by the closure of the pericardium (6-0 C-1 Prolene) and the thorax (5 CCS Stainless steel). Additional analgesia was warranted using bupivacaine (2.5 mg/mL) in the incision after suturing the intramuscular planes (2 CTX plus and 0 MH plus, Vicryl) and prior to suturing the subcutaneous one (2-0 Vicryl); a continuous transdermal suture was performed to close the skin (3-0 PS-1 Monocryl). The chest drains were removed once exudate production was resolved.

Vector production

Capsid-pseudotyped AAV vectors with AAV2 ITRs containing GFP transgene driven by cTnT promoter were produced in 293T cells as previously reported.⁴⁰ AAV vectors were purified by using iodixanol density gradients and buffer exchanged to PBS + 0.001% pluronic using Amicon Ultra 15 mL centrifugal filters (100 kDa; Merck Life Science). Genomic titer was determined by qPCR using double-stranded plasmid DNA templates as standard curves with Fw: CAAAATTTGTGAAAGATTGACTGG; Rv: AAAGCCATACGGGAAGCAAT. AAV vectors in which GFP expression was driven by CMV promoter were provided by Revvity Gene Delivery.

In vitro transduction

NRVMs were isolated from 1- to 2-day-old rats as described previously.⁸⁰ NRVMs were seeded at a density of 4.5 × 10⁵ cells per well in 24-well culture plates coated with a mixture of 12.5% fibronectin and 0.1% gelatin and kept in Tung medium (M199 supplemented with 200 U/L penicillin, 200 μg/L streptomycin, 20 μg/L vitamin B12, 1% MEM non-essential amino acids, 1% HEPES, 3.5 g/L glucose, and 2 mM L-glutamine) with 10% heat-inactivated fetal bovine serum (FBS). The purity of these NRVM preparations is around 74%. The medium was refreshed 1 day after plating and changed to Tung medium with 2% heat-inactivated FBS. On the same day, NRVMs were transduced with AAVs at an MOI of 10⁴ vector genomes (vg)/cell for all groups except AAV9_10×, where an MOI of 10⁵ vg/cell was applied.

Human iPSC-CMs were differentiated as previously reported,⁸¹ seeded in culture plates coated with Matrigel and kept in BPEL medium.⁸² The purity of these hiPSC-CM populations is around 86%. On the same day, cells were transduced with AAVs at an MOI of 5 × 10⁴vg/cell. Medium was refreshed 1 day after transduction and refreshed every 2 days till analyses.

LAASs

Viable LAASs were obtained from patients with atrial fibrillation undergoing thoracoscopic ablation and included in the MARK AF registry (NL5006901819). The left atrial appendage was excised as standard of care during the procedure using a stapling device (Endo Gia stapler, Tyco Healthcare). All patients provided written informed consent.⁸³ Samples were immediately immersed in ice-cold, heparinized (2 mL/L) slicing solution containing 140 mM NaCl, 1.8 mM CaCl₂·H₂O, 4.7 mM KCl, 1 mM MgCl₂·6H₂O, 10 mM glucose, 10 mM HEPES, and 30 mM 2,3-butanedione monoxime (pH 7.4, adjusted with NaOH). The tissue was embedded in 4% low melting point agarose (16520-100, Thermo Fisher Scientific) and mounted on a specimen holder.

LAASs were prepared using a vibrating microtome (7000 smz-2, Campden Instruments).⁸⁴ In a 4°C slicing solution bath, 380-μm slices were cut at 0.02 mm/s with pre-calibrated blades. The slices were then placed in a 37°C recovery solution with the same composition as the slicing solution but without 2,3-butanedione monoxime. Slices were allowed to recover for at least 1 hour before interventions were undertaken.

Prepared human LAASs were cultured in 6-well culture plates with culture inserts (PCHT06H48, Merck Life Science) in M199 medium (11150059, Thermo Fisher Scientific) supplemented with 3% penicillin/streptomycin and 1% ITS supplement (I3146, Merck Life Science). For transduction, 10¹¹ vg of each AAV vector in 50 μL formulation buffer was added onto each slice (corresponding to an MOI of ∼2 × 10⁵ vg/cell). One day post-transduction, LAASs were washed twice with PBS containing 3% penicillin/streptomycin and cultured in M199 medium supplemented with 1% penicillin/streptomycin and 1% ITS supplement. Medium was then refreshed daily until analyses were conducted.

RNA isolation, DNA isolation, cDNA synthesis, and RT-qPCR

Cells were collected in lysis buffer RA1 (Macherey-Nagel). Tissue samples were snap-frozen and preserved at −80°C, followed by homogenization with a SamplePrep CG-200-230 Freezer/Mill Compact Cryogenic Grinder (part #6775, Cole Parmer, Vernon Hills, IL) or kept in RNA preservation buffer immediately after imaging and homogenized using a T10 Basic Ultra Turrax homogenizer (0003737000, IKA-Werke) in lysis buffer RA1. Total RNA and DNA were isolated using NucleoSpin RNA mini kit (Macherey-Nagel) supplemented by DNA/RNA buffer set according to the manufacturer’s protocol. Complementary DNA was transcribed from 200 to 1,000 ng of total RNA with oligo-dT primers (125 μM) and the Superscript II system (Thermo Fisher Scientific). qPCR was performed using the LightCycler 480 Real-Time PCR system (Roche Diagnostics). Relative mRNA expression of the transgene was normalized to that of the reference gene HPRT1 (mouse and human) or ACTB (pig). The number of vg was normalized to the number of diploid target cell genomes based on input DNA assuming 6.5 ng DNA per diploid genome.

Direct fluorescence microscopy

NRVMs

GFP fluorescence and bright field images were taken 3 days post-transduction using an inverted fluorescent microscope (ECLIPSE Ts2R, Nikon Europe, Amstelveen, the Netherlands). Identical exposure and magnification settings were applied for all wells. For hiPSC-CMs, GFP fluorescence images were taken 5 days post-transduction using the same microscope, but images were captured with an external camera with fixed settings.

Mouse tissue

Mice were killed 4 weeks post-AAV injection. GFP fluorescence and bright field images were taken using a fluorescence stereo microscope (M205 FCA, Leica Microsystems) shortly after the heart was isolated. Identical capturing and analysis settings were applied for all animals.

Immunofluorescence staining

Mouse tissue

Tissues were fixed overnight in 4% paraformaldehyde solution in PBS, transferred to 70% ethanol, embedded in Paraplast (Leica Biosystems) and sectioned at 7 μm. Sections were deparaffinized in xylene and rehydrated by an ethanol series of descending concentrations. For antigen retrieval, sections were boiled in unmasking solution (H3300, VectorLabs). Sections were treated with 0.1% Triton X-100 in PBS for 30 min, and non-specific binding sites were blocked using PBS containing 4% bovine serum albumin and incubated overnight with anti-GFP (1:500; ab13970, Abcam), anti-Tnnt2 (1:100; MA5-12960, Thermo Fisher Scientific), and anti-dystrophin (1:150; ab15277, Abcam) antibodies. The next day, the sections were incubated with secondary antibodies conjugated with Alexa Fluor 488, 555, and 647 in combination with the nucleic acid stain DAPI (1 μg/mL; D9542, Merck Life Science). Sections were then mounted, and fluorescence images were acquired using a DM6000 fluorescence microscope (Leica Microsystems).

Pig tissue

Cardiac tissue was snap-frozen and subsequently processed as described above for mouse tissue. Sections were incubated overnight with anti-GFP (1:500; ab13970, Abcam) and anti-Tnnt2 (1:100; MA5-12960, Thermo Fisher) antibodies. The next day, the sections were incubated with secondary antibodies conjugated with Alexa Fluor 488 and 647 in combination with Alexa Fluor 555-conjugated wheat germ agglutinin (1:250, W32464, Thermo Fisher Scientific) and DAPI (1 μg/mL, D9542, Sigma) for visualization of the plasma membrane and cell nucleus, respectively. Sections were then mounted, and fluorescence images were acquired using the aforementioned Leica DM6000 fluorescence microscope.

Picrosirius red staining

Sections were deparaffinized in xylene, rehydrated by an ethanol series of descending concentration, and stained with picrosirius red (P6744, Sigma-Aldrich) to visualize collagen. Stained sections were imaged using a DM5000 fluorescence microscope (Leica Microsystems). Image analysis was performed on 10 non-contiguous sections from the apex of the heart using ImageJ.

Statistical analysis

Statistical analysis was performed using Student’s t test or analysis of variance for repeated measurements (ANOVA) with post-hoc Fisher’s least significant difference (LSD) test in Prism 10 (GraphPad Software, Boston, MA).

Data availability

All data generated in this study are included in this published article and in the supplemental information. All these data will be made available upon reasonable request.

Acknowledgments

This work was supported by European Research Council (starting grant 714866, proof-of-concept grant 899422 and 101081921 to G.J.J.B., proof-of-concept grant 101138069 to J.P.G.S.), Health Holland (LentiPace II to G.J.J.B. and H.L.T.), Horizon 2020 (Eurostars E114245 and E115484 to G.J.J.B. and V.M.C., project TECHNOBEAT, grant number 66724 and RIA-HEAL, grant number 101056712 to J.P.G.S.), Dutch Research Council (Open Technology Program 18485 to H.L.T. and G.J.J.B.; OCENW.GROOT.2019.029 to V.M.C., 2021/TTW/01038252 to G.J.J.B. and J.P.G.S.), European Innovation Council (Pathfinder Challenges Nav1.5-CARED to V.M.C. and G.J.J.B., TRANSITION Project 101099608 TRACTION to V.M.C. and G.J.J.B.), H2020 Marie Sklodowska-Curie Actions COFUND program (No 801540 to J.P.G.S.), and RegMed XB (Cardiovascular Moonshot to A.A.F.d.V.).

Author contributions

Conceptualization, J.W., T.J., V.M.C., and G.J.J.B.; formal analysis, J.W. and T.J.; funding acquisition, G.J.J.B., H.L.T., O.F.K., V.M.C., J.P.G.S., and A.A.F.d.V.; investigation, J.W., T.J., A.C.-B., Z.D., R.N.V., E.E.B., A.R.B., M.K., Y.Y., J.M.V., M.S.J., and T.C.G.; resources, S.S., C.T., C.I.B., A.A.F.d.V.; supervision, G.J.J.B., V.M.C., J.P.G.S., and S.C.A.d.J.; visualization, J.W., T.J., and A.C.-B.; writing—original draft, J.W., T.J., and A.C.-B.; writing—review & editing, all authors.

Declaration of interests

O.F.K., H.L.T., and G.J.J.B. are co-founders of PacingCure BV and report ownership interest in PacingCure BV. J.W. and E.E.B. are employees of PacingCure BV. T.J., V.M.C., and G.J.J.B. have pending patent applications related to this work.